key: cord-0261734-8u73adnk
authors: Li, Jinzhi; Mahoney, Brennan Dale; Jacob, Miles Solomon; Caron, Sophie Jeanne Cécile
title: Visual input into the Drosophila melanogaster mushroom body
date: 2020-05-02
journal: bioRxiv
DOI: 10.1101/2020.02.07.935924
sha: 12970fd08012cac78df72cea4a9cda70582acf0b
doc_id: 261734
cord_uid: 8u73adnk

The ability to integrate input from different sensory systems is a fundamental property of many brains. Yet, the patterns of neuronal connectivity that underlie such multisensory integration remain poorly characterized. The Drosophila melanogaster mushroom body — an associative center required for the formation of olfactory and visual memories — is an ideal system to investigate how different sensory channels converge in higher-order brain centers. The neurons connecting the mushroom body to the olfactory system have been described in great detail, but input from other sensory systems remains poorly defined. Here, we use a range of anatomical and genetic techniques to identify two novel types of mushroom body input neuron that connect visual processing centers — namely the lobula and the posterior lateral protocerebrum — to the dorsal accessory calyx of the mushroom body. Together with previous work that described a pathway conveying visual information from the medulla to the ventral accessory calyx of the mushroom body (Vogt et al., 2016), our study defines a second, parallel pathway that is anatomically poised to convey information from the visual system to the dorsal accessory calyx. This connectivity pattern — the segregation of the visual information into two separate pathways — could be a fundamental feature of the neuronal architecture underlying multisensory integration in associative brain centers.

Sensory systems use different strategies to detect specific physical features of the outside world.

For instance, the olfactory system contains many different types of sensory neuron that are each specialized in detecting a specific class of volatile chemicals. Through only two neuronal layers, 35 olfactory information -the identity of an odor and its concentration -is relayed to higher brain centers (Leinwand and Chalasani, 2011) . In contrast, the visual system contains far fewer types of sensory neuron but, through numerous neuronal layers, it relays a range of highly processed information -for instance, color, brightness, motion and shape -to higher brain centers (Baden et al., 2020) . Thus, higher brain centers have to integrate different types of processed information, 40 bind that information into a coherent representation of the outside world, and use such representations to guide behavior (Yau et al., 2015) . How higher brain centers achieve this feat remains largely unknown. This gap in our knowledge mainly stems from the fact that higher brain centers are formed by a large number of neurons, and that the projection neurons conveying information from different sensory systems to these centers often remain poorly characterized. 45

This makes it difficult to understand whether there are specific patterns of neuronal connectivity that enable multisensory integration and what the nature of these patterns are. Deciphering the fundamental neuronal mechanisms that underlie multisensory integration requires a model system such as the Drosophila melanogaster mushroom body, which consists of a relatively small number of neurons whose connections can be charted reliably. 50

The Drosophila mushroom body is formed by ~2,000 neurons -called the Kenyon cells -and has long been studied for its essential role in the formation of olfactory associative memories (Aso et al., 2014; Hige, 2018) . The identity of the projection neurons that connect the olfactory system to the mushroom body -and the way Kenyon cells integrate input from these 55 neurons -has been characterized in great detail, highlighting fundamental connectivity patterns that enable this higher brain center to represent olfactory information efficiently (Bates et al., 2020; Caron et al., 2013; Tanaka et al., 2012a Tanaka et al., , 2012b Zheng et al., 2018) . Evidence in Drosophila melanogaster shows that the mushroom body is more than an olfactory center, as it is also required for the formation of visual and gustatory associative memories (Liu et al., 1999; Masek 60 and Scott, 2010; Vogt et al., 2014) . However, the identity of the neurons that connect the mushroom body to other sensory systems remains poorly characterized. Thus, a first step towards understanding how the mushroom body integrates multisensory information is to identify such non-olfactory mushroom body input neurons and the genetic tools necessary to manipulate these neurons. 65

The mushroom body receives its input through its calyx and sends its output through its lobes. The calyx -a morphologically distinct neuropil containing the synapses formed between projection neurons and Kenyon cells -can be divided into four, non-overlapping regions: one main calyx as well as three accessory calyces named the dorsal, lateral and ventral accessory 70 calyces (Aso et al., 2014; Yagi et al., 2016) . The five output lobes -the a, a', b, b' and g lobes -contain the synapses formed between Kenyon cells, mushroom body output neurons and dopaminergic neurons (Aso et al., 2014) . With respect to these input and output regions, Kenyon cells can be divided into seven distinct types (Aso et al., 2014) . Of these seven types, five types -the a/bc, a/bs, a'/b'ap, a'/b'm and gmain Kenyon cells -extend their dendrites only into the main 75 calyx and their axons along one or two lobes. Most of the neurons that project to the main calyx emerge from the antennal lobe, the primary olfactory center in the Drosophila brain. Thus a/bc, a/bs, a'/b'ap, a'/b'm and gmain Kenyon cells receive input primarily from the olfactory system (Caron et al., 2013; Aso et al., 2014; Zheng et al., 2018) .

In contrast, the two other classes of Kenyon cells do not extend their dendrites into the main calyx. Instead, the a/bp Kenyon cells extend their dendrites into the dorsal accessory calyx -avoiding completely the main, lateral and ventral accessory calyces -and their axons along the a and b lobes. Likewise, the gd Kenyon cells extend their dendrites exclusively into the ventral accessory calyx and their axons along the g lobe (Aso et al., 2014; Vogt et al., 2016) . Thus, both 85 the a/bp and gd Kenyon cells are anatomically poised to receive non-olfactory input. There is evidence suggesting that the ventral accessory calyx receives input from the medulla, a region of the optic lobe that specializes in processing brightness and color (Morante and Desplan, 2008; Vogt et al., 2016) . Furthermore, a recent study suggests that the dorsal accessory calyx is a multisensory center that integrates input from multiple sensory pathways including the olfactory, 90 gustatory and visual systems (Yagi et al., 2016) .

Here, we report a strategy that uses a combination of genetic tools -including transgenic lines that drive expression in few neurons and a photo-labelling technique used to identify individual neurons and their presynaptic partners -to characterize the input neurons of the a/bp 95 Kenyon cells. We identify two novel types of mushroom body input neuron that, together, form about half of the total input the a/bp Kenyon cells receive in the dorsal accessory calyx. The first neuronal type -henceforth referred to as LOPNs -consists of a neuron that projects from the lobula, a region of the optic lobe specialized in detecting visual features such as shape and motion. The second type of neuron -henceforth referred to as PLPPNs -consists of projection 100 neurons that emerge from the posterior lateral protocerebrum, a brain region that receives input from the optic lobe (Otsuna and Ito, 2006; Keleş and Frye, 2017; Wu et al., 2016) . Interestingly, LOPN and PLPPNs do not project to the ventral accessory calyx and do not connect to the gd Kenyon cells. Based on these findings, we conclude that there are two parallel pathways that convey visual information to the mushroom body: a pathway projecting from the medulla to the gd Kenyon 105 cells and another pathway projecting from the lobula and posterior lateral protocerebrum to the a/bp Kenyon cells.

Neurons projecting to the dorsal accessory calyx emerge from different brain regions

The dorsal accessory calyx is a neuropil formed from the synapses connecting ~90 a/bp Kenyon cells to their input neurons ( Figure 1A ; (Aso et al., 2014) ). Using transgenic lines that drive expression specifically in the a/bp Kenyon cells (the R85D07-GAL4DBD, transgenic line also known as the MB371-GAL4) and transgenic lines that express a photoactivatable form of GFP or PA-GFP (a combination of the UAS-C3PA-GFP and UAS-SPA-GFP transgenic lines, henceforth referred to as UAS-PA-GFP), we photo-labelled individual a/bp Kenyon cells similarly as described in a previous study (Aso et al., 2014; Datta et al., 2008; Ruta et al., 2010) . We found that individual a/bp Kenyon cells extend on average 5 ± 1 claw-shaped 120 dendritic terminals (n = 13) exclusively into the dorsal accessory calyx ( Figure 1B , D) and project their axons along the a and b lobes (not depicted). The overall morphology of a/bp Kenyon cells is similar to the morphology of other types of a/b Kenyon cell; for instance, individual a/bs Kenyon cells extend on average 6 ± 1 claw-shaped dendritic terminals (n = 12) exclusively in the main calyx ( Figure 1C ) and project their axons along the a and b lobes (not depicted). It is worth noting 125 that the claws formed by the a/bp Kenyon cells are much smaller in diameter (1.8 ± 0.4 µm; n = 9 claws) than the claws formed by the a/bs Kenyon cells (3.0 ± 0.4 µm; n = 10 claws; Figure 1B -C (white arrows), E). It is also worth noting that the border of the dorsal accessory calyx is looser than the compact and well-defined circular border of the main calyx. Individual a/bp Kenyon cells can extend dendrites further away from the core of the dorsal accessory calyx, resulting in an 130 irregularly shaped calyx ( Figure 1D) . Thus, the a/bp Kenyon cells resemble a/bs and a/bc Kenyon cells but also differ from them in one major way: the dendrites of the a/bs and a/bc Kenyon cells exclusively innervate the main calyx -a region known to receive most of its input from the olfactory system -whereas dendrites of the a/bp Kenyon cells exclusively innervate the dorsal accessory calyx -a poorly characterized region of the mushroom body. 135

To identify neurons that project to the dorsal accessory calyx and connect to the a/bp Kenyon cells, we used a targeted photo-labeling technique that was adapted from previously published techniques (Aso et al., 2014; Datta et al., 2008; Ruta et al., 2010) . In order to only photo-label the neurons projecting to the dorsal accessory calyx, and not the a/bp Kenyon cells, 140

we used a combination of transgenes that, in concert, drive the expression of PA-GFP in all neurons except the Kenyon cells (the N-Synaptobrevin-GAL4, MB247-GAL80 and UAS-PA-GFP transgenes); instead, Kenyon cells were labeled with the red fluorescent protein DsRed using the MB247-DsRed transgene ( Figure 2A ). The brains of adult flies carrying all the aforementioned transgenes were dissected and imaged using two-photon microscopy ( Figure 2B -C). Guided by 145 the expression of DsRed, we targeted the dorsal accessory calyx -which is clearly distinct from the main calyx -with high-energy light in order to photo-convert PA-GFP specifically in the neurons projecting to that area ( Figure 2C , white dashed outline). Upon photo-labeling, the somata of the neurons that express PA-GFP, and that project either dendrites or axons into the dorsal accessory calyx, were labeled ( Figure 2D -E). On average, 71 ± 22 neurons (n = 22) were 150 photo-labeled. The somata of these neurons are located in seven distinct clusters that are distributed across the brain (Figure 2A , D-E). On the anterior side of the brain, we found three clusters: one located near the antennal lobe (cluster AL containing 3 ± 3 neurons), one located in the optic lobe (cluster OL containing 1 ± 2 neurons) and one located near the anterior ventral lateral protocerebrum (cluster AVLP containing 6 ± 3 neurons) ( Figure 2D , F). On the posterior 155 side of the brain, we found four clusters: one located in the superior medial protocerebrum (cluster SMP containing 4 ± 2 neurons), one located in the superior lateral protocerebrum (cluster SLP containing 20 ± 8 neurons), one located in the lateral horn (cluster LH containing 34 ± 12 neurons) and one located in the posterior lateral protocerebrum (cluster PLP containing 3 ± 2 neurons) ( Figure 2E -F). Altogether, these results suggest that the dorsal accessory calyx receives input 160 from a diverse and distributed collection of projection neurons that can be divided into seven clusters.

While en masse photo-labeling of the neurons projecting to the areas within and around the dorsal accessory calyx gives a good approximation of the number and identity of neurons possibly connecting to the a/bp Kenyon cells, this technique cannot confidently identify true pre-synaptic partners. Additionally, the amount of photo-activated PA-GFP within individual neurons varies 170 significantly between trials, and because a large number of neurons project to the dorsal accessory calyx, it is difficult to resolve the morphology of individual neurons using this technique.

We, therefore, sought to identify GAL4 transgenic lines that drive expression specifically in these neurons. To identify such lines, we carried out an anatomical screen using the FlyLight collection of GAL4 transgenic lines ( Figure 3A ; (Jenett et al., 2012) ). As a first step, we screened through 175 the FlyLight database -an online catalogue that reports the expression patterns of ~7,000 GAL4 driver lines -and selected 267 transgenic lines which that highlighted neuronal processes in the dorsal accessory calyx ( Figure 3B -C).

As a second step, we specifically labeled neurons that project to the dorsal accessory calyx 180 using the same technique as described above but with a different combination of transgenes (the R_line-GAL4, UAS-PA-GFP and MB247-DsRed transgenes). The brains of adult flies that carry this combination of transgenes were dissected and imaged using two-photon microscopy ( Figure 3D , F). Guided by the expression of DsRed, we targeted the dorsal accessory calyx with high-energy light in order to convert PA-GFP specifically in the neurons that project to that region ( Figure 3D , 185 white dashed outline). We screen through the 267 pre-selected transgenic lines and identified 10 lines that we chose to investigate further based on two criteria. First, we selected lines that showed a strong signal after photo-labeling and clear pre-synaptic terminals near or in the dorsal accessory calyx ( Figure 3E , G; Table 1 ). Possible pre-synaptic terminals -bouton-shaped ( Figure 3E , white arrow) -were distinguished from possible post-synaptic terminals -mesh-shaped -based on 190 the recovered photo-labeling signal. Second, we selected lines based on the strength and specificity of their expression patterns: lines expressing at high level in a few neurons led to clear photolabeling signals, making it easier to characterize these putative input neurons further ( Figure 3G ).

Finally, we determined whether the neurons identified with these lines are pre-synaptic 195 partners of the a/bp Kenyon cells. To this end, we used an activity-dependent version of the GFPreconstitution across synaptic partners (GRASP) technique ( Figure 4A ) (Feinberg et al., 2008; Macpherson et al., 2015) . In this technique, the green fluorescent protein (GFP) is split into two complementary fragments -the GFP1-10 and GFP11 fragments -that do not fluoresce when expressed alone; the GFP1-10 fragment is tagged to an anchor that specifically targets the pre-200 synaptic membrane (syb::spGFP1-10), whereas the GFP11 fragment is tagged to an anchor that targets the membrane (spGFP11::CD4). When both fragments are in close contact -in this case, when the syb::spGFP1-10 fragment is released at an active synapse -GFP molecules can be reconstituted and recover their fluorescence. The reconstitution of GFP molecules, as indicated by the presence of fluorescent GFP speckles, suggests that the neurons expressing the two 205 complementary fragments form functional synapses (Macpherson et al., 2015) . In our study, the syb::spGFP1-10 fragment was expressed in the putative input neurons using the different transgenic lines identified in our screen as well as two transgenic lines that a previous study found to drive expression in putative dorsal accessory calyx input neurons ( Figure 4B -M, Table 1 ; (Yagi et al., 2016) ). The expression of the spGFP11::CD4 fragment was driven in most of the Kenyon 210 cells using the MB247-LEXA transgenic line. We look for GFP speckles in the core of the dorsal accessory calyx, which is clearly visible in the red channel as an irregularly shaped structure anterior to the main calyx ( Figure 4B , white dashed outline). We detected a large number of GFP speckles in the core of the dorsal accessory calyx for four of the ten lines that were identified in the screen as well as for the two transgenic lines identified by the 215 4L-M and Table 1 ). Additionally, we detected a small number of GFP speckles in the same region for three other transgenic lines ( Figure 4F -H and Table 1 ). It is worth noting that some of the GFP speckles lay outside of the red signal; most likely these speckles are formed by Kenyon cells that extend dendrites outside of the core of the dorsal accessory calyx ( Figure 4B -H, 4L-M). Finally, we did not detect any GFP speckles for the remaining three lines ( Figure 4I -K and Table 1) . 220

Altogether, our screen identified at least nine transgenic lines that drive expression in neurons that form synapses with the a/bp Kenyon cells of the dorsal accessory calyx.

To characterize further the morphology of the projection neurons that connect to the a/bp Kenyon cells, we photo-labeled individual neurons using the transgenic lines recovered from our screen that showed reliable GRASP signal. To this end, neurons were photo-labeled in the dissected brains of flies carrying the R_line-GAL4 and UAS-PA-GFP transgenes; these brains were then immuno-stained and imaged using confocal microscopy. We first focused our attention on the 230 transgenic lines in which we had found the strongest GRASP signal ( Figure 4B -E and Table 1 ).

Photo-labeling of neurons that project to the dorsal accessory calyx using the R44H11-GAL4 and R72D07-GAL4 transgenic lines revealed a single neuron ( Figure 5A ). In each line, a neuron projecting from the lobula to the dorsal accessory calyx was clearly photo-labeled ( Figure 5B -C). Therefore, we named this type of neuron "LOPN". The soma of LOPN is located in cluster PLP, a 235 region medial to the optic lobe ( Figure 2E , 5D-E). LOPN extends its dendrites to the lobula and projects its axon to the superior lateral protocerebrum and the dorsal accessory calyx, completely avoiding the main calyx and the ventral accessory calyx ( Figure 5D -E). It is worth noting that LOPN is very similar to the OLCT1 neuron that has been previously described using a different transgenic line, R95F09-GAL4; accordingly, we recovered a strong GRASP signal for that 240 transgenic line ( Figure 4L , Table 1 ; (Yagi et al., 2016) ).

We then set out to confirm whether LOPN does indeed connect to a/bp Kenyon cells, as surmised from the GRASP results. A previous study has demonstrated that a technique, which combines photo-labeling and dye-labeling tools, can be used to identify the complement of input 245 that individual Kenyon cells receive from the antennal lobe, and the frequency at which individual projection neurons connect to Kenyon cells (Caron et al., 2013) . We have modified this technique in order to directly measure the connectivity rate between a given projection neuron and Kenyon cells. In this modified version of the technique, we used a combination of transgenic lines to photolabel the projection neuron of interest -using the same protocol that we described in the previous 250 section -and to dye-label a single Kenyon cell ( Figure 6A ). We then assessed how many of the dendritic claws formed by the dye-labeled Kenyon cell are connected to a given projection neuron.

As a proof of principle, we measured the connectivity rate of the DC3 glomerulus projection neurons to the olfactory Kenyon cells that innervate the main calyx. The connectivity rate of the DC3 projection neurons to the olfactory Kenyon cells has been previously approximated to be 255 5.1%; this means that there is a 5.1% chance that a given olfactory Kenyon cell claw receives input from the DC3 projection neuron (Caron et al., 2013) . Using the modified technique, we found that the DC3 projection neurons connect to Kenyon cells at a connectivity rate of 4.3% (n = 30), a value well within the range measured previously (data not shown). We measured the connectivity rates of LOPN to a/bp Kenyon cells using the two transgenes we identified in our 260 screen. Using the R44H11-GAL4 driver line, we found that 4.1 % (n = 27) of the claws formed by the a/bp Kenyon cells connect to LOPN ( Figure 6B -C, H and Table 1 ). Likewise, using the R72H07-GAL4 driver line, we found that 3.0 % (n = 27) of the claws formed by the a/bp Kenyon cells connect to LOPN ( Figure 6H and Table 1 ). Altogether, based on the results obtained using both the GRASP technique and the dye/photo-labeling technique, we conclude that LOPN is a true input 265 neuron of the a/bp Kenyon cells and that it conveys information from the lobula, a visual processing center, to the dorsal accessory calyx.

Photo-labeling of neurons that project to the dorsal accessory calyx using the R19H07-GAL4 and R20G07-GAL4 transgenic lines revealed a group of neurons ( Figure 7A ). In each line, neurons that project from the posterior lateral protocerebrum to the dorsal accessory calyx were photolabeled ( Figure 7B -C). Therefore, we named this type of neuron "PLPPNs". The somata of the PLPPNs are located in cluster LH, a region near the lateral horn ( Figure 7D ). We could identify 13 275 ± 4 PLPPNs (n = 21) per hemisphere using the R19H07-GAL4 transgenic line ( Figure 7D ). The R20G07-GAL4 transgenic line drives expression at a much lower level, making it difficult to analyze the PLPPNs in great detail ( Figure 7C ). PLPPNs extends projections into many brain centers, including the posterior lateral protocerebrum, the superior clamp, the superior lateral protocerebrum and the dorsal accessory calyx, avoiding completely the main calyx and the ventral 280 accessory calyx ( Figure 7D ). Using the dye/photo-labeling technique described above, we measured the connectivity rate between PLPPNs and a/bp Kenyon cells. We found that the PLPPNs identified using the R19H07-GAL4 transgenic line connect at a high rate: we found that 14.1 % (n = 24) of the claws formed by the a/bp Kenyon cells connect to the PLPPNs ( Figure 6D , E, H). We also measured the connectivity rate of individual PLPPNs using the same transgenic line and found 285 that, on average, an a/bp Kenyon cell claw has 1.90% (n = 32) probability of connecting to a PLPPN ( Figure 6H and Table 1 ).

To determine whether the PLPPNs identified in the R19H07-GAL4 and R20G07-GAL4 transgenic lines are the same or different group of neurons, we used the split-GAL4 technique 290 (Luan et al., 2006) . In this technique, the GAL4 transcription factor is split into two complementary fragments -the GAL4AD and GAL4DBD domains -that are both transcriptionally inactive when expressed alone; when both fragments are expressed in the same cell, GAL4 can be reconstituted and recovers its transcriptional activity. PLPPNs were visible in flies that carry the R20G07-GAL4DBD and R19H07-GAL4AD transgenes, thus confirming that both lines are expressed in the 295 same group of PLPPNs ( Figure 7E -G). This result revealed, even more clearly, that individual PLPPNs project in many different brain centers ( Figure 7G ). In order to distinguish the axonal terminals from the dendritic arbors, we used the split-GAL4 combination of transgenic lines (the R20G07-GAL4DBD and R19H07-GAL4AD transgenic lines) to drive the expression of the presynaptic marker synaptotagmin in the PLPPNs ( Figure 7H ). We found that the projections 300 extending into the superior lateral protocerebrum, superior clamp and dorsal accessory calyx contain presynaptic terminals. Similarly, when the expression of the postsynaptic marker DenMark was driven specifically in the PLPPNs, we found that all of the projections made by the PLPPNs contain postsynaptic terminals ( Figure 7I ). However, the projections extending into the posterior lateral protocerebrum are the only projections formed by PLPPNs that contain only post-305 synaptic terminals and no pre-synaptic terminals ( Figure 7H -I).

To identify the neurons that project to the post-synaptic terminals formed by the PLPPNs in the posterior lateral protocerebrum, a poorly characterized visual processing center, we used the targeted photo-labeling technique described above ( Figure 8A ). In short, we used a combination 310 of transgenes that, in concert, drive the expression of PA-GFP in all neurons (the N-Synaptobrevin-QF and QUAS-PA-GFP transgenes) and the expression of tdTomato in the PLPPNs (using the R19H07-GAL4AD, R20G07-GAL4DBD and UAS-tdTomato transgenes). Guided by the expression of tdTomato, we targeted the post-synaptic terminals formed by PLPPNs in the posterior lateral protocerebrum with high-energy light ( Figure 8B , C). Upon photo-labeling, two types of 315 neuron were clearly photo-labeled: the PLPPNs -showing that the photo-activation was specific to these neurons -and neurons projecting from the ventral medulla ( Figure 8D ). These photolabeled neurons project into deeper layers of the medulla ( Figure 8E ). Altogether, from this set of experiments, we conclude that PLPPNs are one of the major input neurons of the a/bp Kenyon cells and that they convey information from the posterior lateral protocerebrum, and possibly from the 320 ventral medulla, to the dorsal accessory calyx.

Photo-labeling using the transgenic lines that displayed weak GRASP signal -these are the 325 R30E11-GAL4, R31C03-GAL4 and R53B06-GAL4 lines -or no GRASP signal -these are the R11F07-GAL4, R11F08-GAL4 and R12C04-GAL4 lines -led to the identification of a third type of neuron (Figure 9 ). The somata of these neurons are all located in AL cluster, a region near the antennal lobe. Although each of these neurons has a distinct overall morphology, they all project from the antennal lobe to the superior lateral protocerebrum and extend their axons in a region 330 near the dorsal accessory calyx. Therefore, we named this type of neuron "ALPNs".

The neurons photo-labeled using the R30E11-GAL4 and R31C03-GAL4 driver lines show a nearly identical morphology: in each line, a single neuron that extends its dendrites into the posterior antennal lobe and project its axons into the superior lateral protocerebrum was visible 335 ( Figure 9A ). We named this neuron "ALPN1". It is worth noting that ALPN1 is very similar to the thermosensitive AC neuron that has been previously described (Shih and Chiang, 2011) . Using a combination of photo-labeling and dye-tracing techniques, as described above, we measured the connectivity rate between ALPN1 and a/bp Kenyon cells. We could not detect any connections between ALPN1 and a/bp Kenyon cells (n = 26), suggesting that ALPN1 is not a dorsal accessory 340 calyx input neuron ( Figure 6H , Table 1 ). Similarly, the neuron photo-labeled using the R53B06-GAL4 line extends its dendrites into the column region of the posterior antennal lobe, a region known to be activated by high humidity, and the sub-esophageal ganglion, a gustatory processing center ( Figure 9B ). We named this neuron "ALPN2". Again, we could not detect any connections between ALPN2 and a/bp Kenyon cells (n = 30), suggesting that ALPN2 is not a dorsal accessory 345 calyx input neuron ( Figure 6F -H, Table 1 ). Thus, ALPN1 and ALPN2 are most likely not contributing major input to the a/bp Kenyon cells.

We extended our analysis to the transgenic lines that displayed no GRASP signal. The neurons photo-labeled using the R11F08-GAL4 and R12C04-GAL4 driver lines show an overall 350 similar morphology: in each line, a single neuron that extends its dendrites into the arm region of the posterior antennal lobe, a region known to be activated by low humidity, was visible ( Figure   9D ). We named this neuron "ALPN3". Not surprisingly, we could not detect any connections between ALPN3 and a/bp Kenyon cells (n = 26) suggesting that ALPN3 does not provide input into the dorsal accessory calyx (Table 1) . Similarly, the neuron photo-labeled using the R11F07-GAL4 355 line extends its dendrites broadly throughout the anterior antennal lobe, an olfactory processing center ( Figure 9F ). We named this neuron "ALPN4". Again, we could not detect any connections between ALPN4 and a/bp Kenyon cells (n = 25), suggesting that ALPN4 does not provide input into the dorsal accessory calyx (Table 1) . Thus, we could confirm that the transgenic lines that displayed no GRASP signal do not appear to provide major input to the a/bp Kenyon cells. 360

Altogether, these results show that the third type of neuron identified in our screen -ALPNsproject to a region close to the dorsal accessory calyx but are most likely not presynaptic to the a/bp Kenyon cells.

Our results suggest that the LOPN and PLPPNs that we identified in our screen connect to the a/bp Kenyon cells and that, together, these two types of projection neuron represent a large fraction of the input neurons of the dorsal accessory calyx. Our results also suggest that the ALPNs do not connect to the a/bp Kenyon cells, as another study suggested (Yagi et al., 2016) . Thus, we 370 conclude that the dorsal accessory calyx is anatomically poised to receive information primarily from the lobula and the posterior lateral protocerebrum, two visual processing centers. We verified whether these findings corroborate with the recently released Drosophila hemibrain connectome (Xu et al., 2020) . Using the Neuprint 1.0.1 platform, we focused our attention on the 60 reconstructed a/bp Kenyon cells that have been fully traced. Altogether, these a/bp Kenyon cells 375 receive input from a large number of neurons but most of these synapses are most likely axoaxonic synapses as they are located along the a and b lobes. These input neurons are primarily other Kenyon cells, various dopaminergic neurons as well as neurons known to connect broadly to all Kenyon cells such as the APL neuron and the DPM neuron (data not shown). We focused our attention on the 141 input neurons that connect to the a/bp Kenyon cells in the dorsal 380 accessory calyx. Not surprisingly, in accordance with the results obtained from our en masse photo-labeling experiment, the cell bodies of these input neurons can be divided into seven different clusters and the ratio of neurons belonging to a given cluster is largely consistent between both analyses ( Figure 10A ).

Interestingly, the LH cluster is the most numerous cluster: our study identified 34 ± 12 neurons (n = 22) in this cluster -including the 13 ± 4 PLPPNs (n = 21) -whereas the connectome shows a total of 61 input neurons with cell bodies located in the lateral horn ( Figure 10A ). Among these neurons, we identified 13 that are morphologically highly similar to PLPPNs ( Figure 7J -K, Table 2 ). We found that these PLPPN-like neurons connect to 59 of the 60 reconstructed a/bp 390 Kenyon cells, representing 49.40% of the input all a/bp Kenyon cells receive in the dorsal accessory calyx ( Figure 10B and Table 2 ). PLPPNs-like neurons connect mostly to a/bp Kenyon cells but also form a few connections to gd and gmain Kenyon cells, reinforcing the observation that the dorsal accessory calyx input neurons and the ventral accessory calyx input neurons form two parallel pathways ( Figure 10C ). In our study, we identified neurons connecting the ventral medulla 395 to the PLPPNs (Figure 8 ) but, because the hemibrain connectome does not include the medulla, we could not identify fully traced neurons similar to these PLPPNs input neurons. Additionally, our study identified 1 ± 2 neurons (n = 22) in the OL cluster -including LOPN -whereas the connectome shows a total of 4 input neurons with cell bodies located in the optic lobe ( Figure   10A ). Among these neurons, we identified a neuron very similar to LOPN (Figure 5F -G). We found 400 that this LOPN-like neuron connects to 11 of the 60 reconstructed a/bp Kenyon cells, representing 2.66% of the input all a/bp Kenyon cells receive in the dorsal accessory calyx ( Figure 10B and Table 2 ). As we found in our study, this LOPN-like neuron does not connect to any other types of Kenyon cell, forming a pathway parallel to the one conveying visual information to the ventral accessory calyx. 405

Finally, both studies identified a number of neurons with cell bodies located near the antennal lobe: using en masse photo-labeling, we identified a total of 3 ± 3 neurons (n = 22) in the AL cluster, whereas nine such neurons are found in the connectome. Among these neurons, we recognized three of the four ALPNs we characterized: an ALPN2-like neuron and an ALPN4-like 410 neuron that represents, respectively, 0.39% and 0.54% of the input that a/bp Kenyon cells receive in the dorsal accessory calyx ( Figure 9C , G and Table 2 ). We could not recognize an ALPN1-like neuron in the connectome but we identified an ALPN3-like neuron ( Figure 9E , Table 2 ). The ALPN3like neuron does not connect to any Kenyon cells. Together, the nine antennal lobe input neurons -including the ALPN2-like and ALPN4-like neurons -represent less than 2% of the input that 415 a/bp Kenyon cells receive in the dorsal accessory calyx ( Figure 10B ). Interestingly, most of the remaining input a/bp Kenyon cells receive in the dorsal accessory calyx (22.93% of the total input a/bp Kenyon cells receive in the dorsal accessory calyx) is from a single neuron projecting from many brain regions including the main calyx and the pedunculus, two regions of the mushroom body, as well as the superior lateral protocerebrum (Table 2) . Altogether, these observations 420 confirm that LOPN-like and PLPPN-like neurons are the major input neurons that connect to the a/bp Kenyon cells in the dorsal accessory calyx. Thus, the dorsal accessory calyx receives most of its input from visual processing centers -the posterior lateral protocerebrum and the lobula -and is thus anatomically poised to process mostly, if not strictly, visual information.

In this study, we identified and characterized neurons projecting to the dorsal accessory calyx of the mushroom body and show that these neurons are presynaptic to the a/bp Kenyon cells ( Figure   11 ). Using a combination of genetic and anatomical techniques, we could distinguish two different 430 types of projection neuron: LOPN projecting from the lobula -an area of the optic lobe processing visual features such as shape and motion -and the PLPPNs projecting from the posterior lateral protocerebrum. Although the posterior lateral protocerebrum remains poorly characterized in D.

melanogaster, evidence from other insects shows that this brain region receives input from the optic lobe (Paulk et al., 2008 (Paulk et al., , 2009 ). Interestingly, we found that the dendrites formed by the 435 PLPPNs in the posterior lateral protocerebrum are in close proximity to neurons that project from the ventral medulla. Based on our results -and considering insights from the connectome -we estimate that LOPNs and PLPPNs account for half of total input that a/bp Kenyon cells receive in the dorsal accessory calyx. LOPNs and PLPPNs do not extend axonal terminals into the ventral accessory calyx, the other calyx known to receive visual input, but rather extend axonal terminals 440 into the dorsal accessory calyx and into the superior lateral protocerebrum. Likewise, the a/bp Kenyon cells do not connect to the visual projection neurons that are associated with the ventral accessory calyx (Vogt et al., 2016) . These findings suggest that the visual system is connected to the mushroom body via two parallel pathways: the a/bp Kenyon cells receive input from the lobula and the posterior lateral protocerebrum, whereas the gd Kenyon cells receive input directly 445 from the medulla. Further functional studies are necessary to determine what kind of visual information is processed by the a/bp Kenyon cells.

In Drosophila melanogaster, the mushroom body has long been studied as an olfactory processing center. However, evidence from many insects, including the honeybee Apis mellifera, 450

shows that the mushroom body integrates sensory information across different modalities. In honeybees, the input region of the mushroom body, also called the calyx, is divided into different layers and each layer receives input from either the olfactory or visual system (Gronenberg, 2001) . Because the dendrites of Kenyon cells are also restricted to specific layers, it has been suggested that, in the honeybee, multisensory integration does not occur at the level of individual 455

Kenyon cells, but rather at the population level (Ehmer and Gronenberg, 2002) . Although the honeybee mushroom body differs greatly from the Drosophila mushroom body -it contains about a hundred times as many Kenyon cells and its input region is divided in multiple complex layers -it appears that both mushroom bodies share a common fundamental connectivity principle: the segregation of input based on sensory modality. This connectivity mechanism is immediately 460 apparent in the structural organization of the Drosophila melanogaster mushroom body: the Kenyon cells receiving input from the olfactory system all extend their dendrites into the main calyx, whereas as the Kenyon cells receiving input from the visual system extend their dendrites either in the dorsal accessory calyx or the ventral accessory calyx. Many studies have demonstrated that the Kenyon cells that process olfactory information -those associated with 465 the main calyx -integrate input broadly across the different types of olfactory projection neuron (Caron et al., 2013; Zheng et al., 2018) . Interestingly, it appears that the Kenyon cells that process visual information are wired differently.

We have a thorough understanding of how olfactory Kenyon cells integrate input from the 470 antennal lobe: most Kenyon cells receive, on average, input from seven projection neurons and the projection neurons connecting to the same Kenyon cell share no apparent common features (Caron et al., 2013; Zheng et al., 2018) . Theoretical studies have shown that this random-like connectivity pattern enables the mushroom body to form sparse and decorrelated odor representations and thus maximizes learning (Litwin-Kumar et al., 2017) . Randomization of 475 sensory input is a connectivity pattern that is well suited for representing olfactory informationas an odor is encoded based on the ensemble of olfactory receptors it activates -and might not be suitable for representing visual information. Indeed, our results suggest that that specific visual features -the signals processed by the medulla and the ones processed by the lobula and the posterior lateral protocerebrum -need to be represented by two separate subpopulations of 480 Kenyon cells. This observation mirrors anatomical studies of the honeybee brain: the neurons projecting from the lobula terminate in a different layer than the neurons projecting from the medulla (Ehmer and Gronenberg, 2002) . This arrangement might be essential to preserve distinct visual features when forming associative memories. Functional and behavioral studies are required to determine whether indeed the mushroom body represents multisensory stimuli in this 485 manner.

Flies (Drosophila melanogaster) were raised on standard cornmeal agar medium and maintained in an incubator set at 25°C, 60% humidity with a 12 hours light / 12 hours dark cycle (Percival Scientific, Inc.). Crosses were set up and reared under the same conditions, but the standard cornmeal agar medium was supplemented with dry yeast. The strains used in this study are 495 described in the table below. 

Two to six-day-old flies were used for all photo-labeling experiments. The protocol was largely based on the one developed in a previous study (Aso et al., 2014) . In short, brains were dissected in saline (108 mM NaCl, 5 mM KCl, 5 mM HEPES, 5 mM Trehalose, 10 mM Sucrose, 1 mM NaH2PO4, 4 mM NaHCO3, 2 mM CaCl2, 4 mM MgCl2, pH≈7.3), treated for 1 minute with 2 mg/ml 505 collagenase (Sigma-Aldrich) and mounted on a piece of Sylgard placed at the bottom of a Petri dish. Photo-labeling and image acquisition were performed using an Ultima two-photon laser scanning microscope (Bruker) with an ultrafast Chameleon Ti:Saphirre laser (Coherent) modulated by Pockels Cells (Conotopics). For photo-labeling, the laser was tuned to 710 nm with an intensity of 5-30 mW; for image acquisition, the laser was tuned to 925 nm with an intensity of 510 1-14 mW (both power values were measured behind the objective lens). A 60X water-immersion objective lens (Olympus) was used for both photo-labeling and image acquisition. A 40X waterimmersion objective lens (Olympus) was used for image acquisition in some experiments (the ones described in Figure 2 and Figure 3) . A GaAsP detector (Hamamatsu Photonics) and PMT detector were used for measuring green and red fluorescence, respectively. Photo-labeling and 515 image acquisition files were visualized on a computer using the Prairie View software version 5.4 (Bruker). Image acquisition was performed at a resolution of 512 by 512 pixels, with a pixel size of 0.39 μm (60X lens) or 0.582 μm (40X) and a pixel dwell time of 4 μs. Each pixel was scanned 8 times.

For photo-labeling of the dorsal accessory calyx (or the posterior lateral protocerebrum), a volume spanning the entire neuropil was divided into eight to 12 planes with a step size of 2 μm. The mask function of the Prairie View software was used to mark the targeted region in every plane, and the boundaries of the mask were determined based on the red fluorescent protein DsRed expressed by the a/bp Kenyon cells. The photo-labeling step was performed using a pixel 525 size of 0.019 μm and a pixel dwell time of 2 μs. Each pixel was scanned four times. Each plane was scanned 30 times with 30-second interval. The entire photo-labeling cycle was repeated two to three times, with a 10-minute resting period between cycles. The entire brain was imaged before and after photo-labeling using the 40X water-immersion objective lens (Olympus, Japan).

The number of cell bodies recovered after the photo-labeling was measured using the Multi-Point 530

Tool function of the Fiji software (Schindelin et al., 2012) .

For the photo-labeling of single neurons, a slightly modified protocol was used. Instead of a volume, a single square plane of 1.0 μm by 1.0 μm centered on the soma of a neuron cell was scanned 70 to 100 times with a 10 seconds interval between scans. Photo-labeled brains were 535 fixed in 2% paraformaldehyde diluted in 1X phosphate bovine saline (PFA) for 45 minutes at room temperature, washed five times in 0.3% Triton X-100 diluted in 1X phosphate bovine saline Technologies) were used. On the following day, brains were washed four times in PBST and mounted on a slide (Fisher Scientific) using the mounting media VECTASHIELD (Vector Laboratories Inc.). Immuno-stained brains were imaged using an LSM 880 confocal microscope.

(Zeiss). The neuropils innervated by the input neurons were identified by comparing the confocal images with the adult brain template JFRC2 available on Virtual Fly Brain 565 (https://v2.virtualflybrain.org/) (Jenett et al., 2012) .

One to three-day-old flies were used when mapping the connectivity rate between input neurons 570 and α/βp Kenyon cells. Brains were dissected in saline, treated for one minute with 2 mg/ml collagenase (Sigma-Aldrich) and mounted on a piece of Sylgard placed at the bottom of a Petri dish. The imaging protocol is the same as described above but the photo-labeling protocol is different. Each of the input neurons was photo-labeled using a single plane centered on either its soma or its projection and by scanning the plane three to five times. Each pixel was scanned eight 575 times with a pixel size of 0.019 μm and a pixel dwell time of 4 μs. A fire-polished borosilicate glass pipette (0.5 mm I.D., 1.0 mm O.D., 10 cm length; Sutter Instruments) was pulled using the P-2000 micropipette puller (Sutter Instruments) and backfilled with Texas Red dye (lysine-fixable 3000 MW;

Life Technologies) dissolved in saline. The tip of the pipette was positioned next to the cell body of a randomly chosen α/βp Kenyon cell under the two-photon microscope. The dye was electroporated 580 into the cell body using three to five 1-5 millisecond pulses of 20-50 V. The dye was allowed to diffuse within the Kenyon cell for 5 minutes before the brain was imaged.

All confocal images were collected on LSM880 confocal microscope (Zeiss). For imaging whole brain, each sample was imaged twice using a Plan-Apochromat 40X/1.3 Oil M27 objective lens.

First, the entire brain was divided into four tiles, each tile was imaged separately (voxel size = 0.46 μm by 0.46 μm by 2 μm, 764 by 764 pixels per image plane) and then stitched together using the Stitch function of the Zen microscope software (Zeiss). For imaging specific neuropil regions, 590 the same objective lens was used, but with higher resolution (voxel size = 0.09 μm by 0.09 μm by 1 μm, 2320 by 2320 pixels per image plane). For imaging brains manipulated using the GRASP technique, a Plan-Apochromat 63X/1.4 Oil M27 objective lens was used in combination with the RGB Airyscan mode. Images were processed using the Airyscan function of the Zen microscope software. All confocal images were analyzed using the Fiji software (Schindelin et al., 2012) . All 595 figure panels are maximum intensity projection of confocal stacks or sub-stacks. (A) A schematic of the Drosophila brain shows how the frequency of connections between the α/βp Kenyon cells (red) and a given input neuron (green), here LOPN, was measured. (B-G) A 905

given input neuron was photo-labeled (bright green) and a randomly chosen a/bp Kenyon cell was dye-filled (red). The total number of claws formed by the dye-filled Kenyon cell was counted and claws connecting to the axonal terminals of the photo-labeled neuron were detected (arrows).

Such connections were found for LOPN (B-C), PLPPNs (D-E) but not for the other neurons identified in this study such as ALPN2 (F-G). (H) The frequency of connections between α/βp Kenyon cells 910 and a given input neuron was calculated by dividing the number of connections detected (for instance arrows in C and E) by the total number of claws sampled for that particular input neuron.

The following genotypes were used in this figure: yw/yw;UAS-C3PA-GFP unknown /UAS-SPA-GFP attP40 ; R13F02-GAL4AD VK00027 ,R85D07-GAL4DBD attP2 /R_line(as indicated in the panel)- 

The neuronal architecture of the mushroom body provides a logic for associative learning

Understanding the retinal basis of vision across species

Complete connectomic reconstruction of olfactory projection neurons in the fly brain

Random convergence of olfactory inputs in the Drosophila mushroom body

The Drosophila pheromone cVA activates a sexually dimorphic neural circuit

Segregation of visual input to the mushroom bodies in the 635 honeybee (Apis mellifera)

GFP Reconstitution Across Synaptic Partners (GRASP) Defines Cell Contacts and 640 Synapses in Living Nervous Systems

Subdivisions of hymenopteran mushroom body calyces by their afferent supply

Systematic analysis of the visual projection neurons of Drosophila melanogaster: Lobula-specific pathways

What can tiny mushrooms in fruit flies tell us about learning and memory? 650

A GAL4-Driver Line Resource for Drosophila Neurobiology

Object-Detecting Neurons in Drosophila

Olfactory networks: From sensation to perception

Context generalization in Drosophila visual learning requires the mushroom bodies

Refined Spatial Manipulation of Neuronal Function by Combinatorial Restriction of Transgene Expression

Dynamic labelling of neural connections in multiple colours by trans-synaptic fluorescence complementation

Connectomics analysis reveals first, second, and third order thermosensory and hygrosensory neurons in the adult Drosophila brain

Limited taste discrimination in Drosophila

The Color-Vision Circuit in the Medulla of Drosophila

Visual processing in the central bee brain

The processing of 695 color, motion, and stimulus timing are anatomically segregated in the bumblebee brain

A dimorphic pheromone circuit in Drosophila from sensory input to descending output

Fiji: An open-source platform for biological-image analysis. 705

Organization of antennal lobe-associated neurons in adult Drosophila melanogaster brain

Dye fills reveal additional olfactory tracts in the protocerebrum of wild-type Drosophila

Direct neural pathways convey distinct visual information to drosophila mushroom bodies

Shared 720 mushroom body circuits underlie visual and olfactory memories in Drosophila

Visual projection neurons in the Drosophila lobula link feature detection to distinct 725 behavioral programs

A Connectome of the Adult Drosophila Central Brain

Convergence of multimodal sensory pathways to the mushroom body calyx in Drosophila melanogaster

Dissecting neural circuits for multisensory integration and crossmodal processing

A naturally monomeric infrared fluorescent protein for protein labeling in vivo

A Complete Electron Microscopy Volume of the Brain of Adult Drosophila 760 melanogaster

R19H07-GAL4) or weakly (C, R44H11-GAL4) in neurons projecting to or from the dorsal accessory calyx (white dashed outline) were selected for further investigation. These images were obtained from the FlyLight website. (D, E) The dorsal accessory calyx (white 850 dashed outline) of the mushroom body (red) was visualized in the selected lines, here R19H07-GAL4, and targeted for photo-activation by designing a mask that exposed the outlined region to high energy light; transgenic lines driving expression in a few neurons extending clear axonal terminals in the dorsal accessory calyx (E, arrow) were selected for further investigation. (F, G) Before photo-labeling

The following genotype was used in the D-G panels: yw/yw;MB247-DsRed unknown ,UAS-C3PA-GFP unknown /UAS-C3PA-GFP attP40

B-C) LOPN (bright green) was identified in the screen using two different transgenic lines (B: R44H11-GAL4 and C: R72D07-GAL4); the neurons photo-885 labeled in each line show an overall similar morphology. (D-E) LOPN was photo-labeled using either the R44H11-GAL4 (D) R72D07-GAL4 (E) transgenic lines; the samples were fixed, immuno-stained (nc82 antibody, blue) and imaged. The photo-labeled neurons show an overall similar morphology: their somata are located near the optic lobe; they extend dendritic terminals in a small region of the lobula (LO); they extend axonal terminals in the dorsal accessory calyx 890 (dACA) and the superior lateral protocerebrum (SLP). (F-G) A LOPN-like neuron was identified in the hemibrain connectome: (F) this neuron projects from the lobula to the dorsal accessory calyx and the superior lateral protocerebrum; (G) its axonal terminals innervate the dorsal accessory calyx (dACA), but not the main calyx (CA) or the ventral accessory calyx (vACA)

E) yw/yw;MB247-DsRed unknown ,UAS-C3PA-GFP attP40 /UAS-C3PA-GFP unknown

B-C) A group of PLPPNs (bright green) were identified in the screen using two different transgenic lines (B: R19H07-GAL4 and C: R20G07-GAL4). (D) PLPPNs were photolabeled using the R19H07-GAL4 transgenic line; the sample was fixed, immuno-stained (nc82 antibody, blue) and imaged. The photo-labeled neurons show an overall similar morphology: their 925 somata are located near the lateral horn; they extend projections in the posterior lateral protocerebrum (PLP), the superior clamp (SC), the superior lateral protocerebrum (SLP) and the dorsal accessory calyx (dACA). (E-F) The expression patterns observed for each of the GAL4 lines are broad and include many neurons; (E) the R19H07-GAL4 line drives expression strongly in many neurons, including the PLPPNs, (F) whereas the R20G07-GAL4 line drives expression 930 weakly in fewer neurons, including the PLPPNs. These images were obtained from the FlyLight website. (G) A split-GAL4

J) these neurons project from the posterior lateral protocerebrum to the dorsal accessory calyx, the superior lateral protocerebrum and the superior clamp; (K) their axonal terminals innervate the dorsal accessory calyx (dACA), but not the main calyx (CA) or the ventral accessory calyx (vACA). The following genotypes were used in this figure: (B, D) yw/yw; MB247-DsRed unknown , UAS-C3PA-940 GFP unknown /UAS-C3PA-GFP attP40

C) yw/yw; UAS-C3PA-GFP unknown /CyO;UAS-C3PA-GFP attP2 , R20G07-GAL4 attP2 /MKRS; (G) yw/yw; MB247-DsRed unknown , UAS-C3PA-GFP unknown /CyO; UAS-C3PA-GFP attP2

unknown ; (I) yw/yw;UAS-DenMark 2 /CyO

The authors declare that no competing interests exist.